All News

A University of Maryland computational biologist has received seed funding for an interdisciplinary project that uses machine learning to help untangle the myriad of causes behind hearing loss.

Michael Cummings, a professor of biology with an appointment in the University of Maryland Institute for Advanced Computer Studies, is co-PI on a $150,000 grant from the UMD Brain and Behavior Institute (BBI).

The project was one of five that received BBI seed funding this year as part of an interdisciplinary program focused on generating novel tools and approaches to understand complex behaviors produced by the human brain.

Cummings, who is also director of the Center for Bioinformatics and Computational Biology, will collaborate on the project with Matthew Goupell, a professor in the Department of Hearing and Speech Sciences and director of the Auditory Perception and Modeling Lab.

“I’m thankful for this opportunity to further our understanding of age-related hearing loss, and for this seed grant that makes our initial collaboration with the Goupell lab possible," Cummings says.

The UMD team is focused on the causes of sensory and cognitive impairment that can rapidly increase with age.

Up to 40 percent of individuals older than 70 suffer from hearing and vision loss, Cummings says, and without the ability to communicate effectively, can become reclusive, frustrated and depressed.

In addition, about 25 percent of Americans older than 65 have some mild cognitive decline and 10 percent have Alzheimer’s disease.

Although each of these impairments have been studied individually, much less is known about the relationship between hearing, mild cognitive impairment, and dementia due to Alzheimer’s disease or other causes, except that there is growing evidence of a link.

To understand how aging affects our sensory and cognitive systems, the UMD team believes it imperative to decipher the joint role of hearing and speech understanding ability, the neural mechanisms that contribute to age-related declines in these areas, and what role hearing plays in age-related cognitive impairment and dementia due to Alzheimer’s disease or other causes.

Their study will use machine learning on a substantial existing database of behavioral, clinical and cognitive data that have been previously collected to generate results for subsequent research.

“The problem is very complex when you consider the relationship between aging, sensation and cognition in individual people,” Goupell says. “While we have known for years that there is a correlation between hearing loss and cognitive decline, we have not made much progress on understanding why this occurs. Machine learning will help guide us to the most probable answers, particularly ones that we may not have considered yet, which will allow us to design future studies with strong testable hypotheses.”

—Story by Melissa Brachfeld

Jamshed Khan, a third-year computer science doctoral student, recently received the Ian Lawson Van Toch Memorial Award for Outstanding Student Paper at the 2021 Conference on Intelligent Systems for Molecular Biology (ISMB) and European Conference on Computational Biology (ECCB).

ISMB is the flagship meeting of the International Society for Computational Biology (ISCB). The event is co-located with ECCB and has grown to become the world’s largest bioinformatics and computational biology conference. The conference was held virtually July 25–30 due to the ongoing COVID-19 pandemic.

The award went to “Cuttlefish: Fast, parallel, and low-memory compaction of de Bruijn graphs from large-scale genome collections,” authored by Khan and his adviser Rob Patro, an associate professor of computer science with an appointment in the University of Maryland Institute for Advanced Computer Studies (UMIACS). The paper introduces a new algorithm for efficiently building a compacted de Bruijn graph from collections of input genome or transcriptome sequences.

The de Bruijn graph and its colored compacted variant has become an increasingly useful data structure in genomic analysis. The structure is used to assemble new genomes, compare the genomes of related organisms, and build efficient indexes for mapping or aligning the data from sequencing experiments, among other tasks.

However, the process of building the compacted de Bruijn graph is a computational challenge, with many existing solutions requiring a lot of time and memory as the input becomes larger.

In the paper, Khan and Patro—who are both members of the Center for Bioinformatics and Computational Biology (CBCB)—present a highly scalable and very low-memory algorithm called Cuttlefish, to construct the compacted de Bruijn graph for collections of whole genome references.

The researchers say Cuttlefish considerably outperforms existing state-of-the-art approaches. It can build the compacted de Bruijn graph for 100 human genomes two and half times as fast and using less than a quarter of the memory compared to the next best tool performing the same task.

“By providing a faster and more memory-frugal algorithm for constructing this widely-used structure of the compacted de Bruijn graph, we anticipate that Cuttlefish will aid in some of the many downstream applications in which this graph is used, like comparative genomics, pangenome analysis, and the indexing of large collections of related genomes for sequence analysis,” Patro says.

The Ian Lawson Van Toch Memorial Award for Outstanding Student Paper is given to the student who presents the most thought-provoking or original paper at ISMB/ECCB, as judged by a panel of experts. The award, which is sponsored by the Princess Margaret Hospital Foundation, is given in memory of Ian Lawson Van Toch, a 23-year-old medical biophysics graduate student at the University of Toronto who passed away in August 2007.

Khan says he is honored to receive such a prestigious award.

“We are ecstatic for this recognition of our work,” he says. “I'm very grateful to Rob for his support and guidance along the way, and for being such a remarkable adviser and mentor. I'm also thankful to my talented and dedicated lab-mates for their critiques and feedback. We look forward to Cuttlefish paving the way toward more exciting results in high-throughput genomics research.”

Theresa Alexander, a graduate student in the Center for Bioinformatics and Computational Biology (CBCB), loves all kinds of creatures and spends her free time with her dog Fibonacci and riding her horses Ella, Spotty and Fergus.

As a fourth-year doctoral student in biological sciences, Alexander’s love of animals extends down to investigating very specific biological problems that affect them.

She earned her master’s degree in epidemiology and biostatistics with a focus on genetic biostatistics from Case Western Reserve University in 2017. Currently, she is co-advised by Najib El-Sayed, a professor of cell biology and molecular genetics with a joint appointment in the University of Maryland Institute for Advanced Computer Studies (UMIACS), and Colenso Speer, an assistant professor of biology.

Her most recent work focuses on investigating a specific neuronal cell type that syncs mammals' sleep/wake cycle—known as circadian rhythm—to the sun cycle. These specific cells in the brain—called intrinsically photosensitive retinal ganglion cells (ipRGCs)—have a protein that makes them photosensitive, meaning these neurons send action potentials from your eye to a specific part of the brain when light photons enter your eye.

Alexander says neurons have to package and ship mRNAs far distances to then be turned into proteins in these distal parts of the cell. Messenger ribonucleic acid—or mRNA for short—is a molecule that carries genetic code from DNA to the rest of the cell.

“One problem we are trying to understand is what mRNAs actually get shipped out to these far-away compartments in these specific neurons—i.e. how does the USPS sorting center in the neuron cell body decide which mRNAs get packaged up and sent out to the processing centers in the ‘rural’ parts of the cell,” she explains.

The second major problem involves looking at the heterogeneity between different neuronal cell-types and sub-cell types. IpRGCs can be split into six different “sub-cell” types. Each of these cell types goes from the retina to different brain targets to perform different functions—some help us maintain our sleep-wake cycles in coordination with the sun cycle while others have image-forming visual functions.

“What we want to help shed light on is how these cells differ from other subtypes based on which genes each expresses throughout development at different critical neuronal growth stages,” Alexander says.

Her group is currently studying ipRGCs in mice, but says she hopes that these results extend to ipRGC function in other mammals.

Alexander enjoys working in CBCB because of the sense of community and the wide array of research being conducted.

“It is absolutely amazing to get a room full of people from so many backgrounds and specialties together,” she says. “There's always a unique perspective to solve a problem, and if I am working on a new problem or an area which I am not familiar with, there's always someone to go to who has some experience with it.”

El-Sayed says Alexander contributes a “wealth of interdisciplinary knowledge” to his lab as well as CBCB.

“Theresa is well-grounded both in the life sciences and bioinformatics and brings a deep understanding of statistical inference to our analyses,” he says. “In addition to her impressive intellectual capacity, she is pleasant, positive and simply fun to work with.”

Alexander is also a UMD COMBINE (Computation and Mathematics for Biological Networks) fellow, a National Science Foundation-funded Research Traineeship (NRT) program in network biology.

Additionally, she volunteers for "Girls Talk Math,” a summer program for high school students that is run by UMD’s math department. The program exposes students to math topics that they may not come across in their normal curriculum. Alexander teaches network science.

After she earns her doctorate, Alexander envisions applying her data science skills to biotech research.

“The intersection of biology, computer science and statistics is the most exciting place to me, and goal-oriented industry research seems like a perfect place for that,” she says.

—Story by Melissa Brachfeld

Perhaps the most iconic image in evolutionary biology is Charles Darwin's sketch of an evolutionary tree. The illustration highlights Darwin’s transformational idea that the evolutionary relationships among species can be depicted through a branching pattern, a concept known as the Tree of Life.

While Darwin primarily relied on the physical characteristics shared by subsets of species to determine an evolutionary tree’s structure, scientists today are using vast amounts of genomic data to reconstruct evolutionary trees—a field known as phylogenetics.

Erin Molloy, who joins the University of Maryland on July 1 as an assistant professor in the Department of Computer Science, is part of this new genomic revolution.

She is using powerful computational tools to unlock the full breadth of information available in genomic data, designing efficient algorithms to estimate evolutionary trees. This type of information is vital for determining the evolutionary history of birds, plants and even microbes, such as SARS-CoV-2, the virus which causes COVID-19.

“The main goal is to estimate the tree—and other parameters—given the observed genomic data,” says Molloy, who will also hold an appointment in the University of Maryland Institute for Advanced Computer Studies (UMIACS). “The resulting phylogeny is not only interesting in its own right, but it is also important for downstream analyses.”

For example, she notes that the phylogeny for SARS-CoV-2 is not only useful for studying how the virus evolves and how new strains emerge, but also for strain identification—that is, determining which strains are present in a sample—and even contact tracing.

Molloy recently completed a year-long postdoctoral researcher position in the Machine Learning and Genomics Lab at the University of California, Los Angeles.

She says she is looking forward to expanding her research agenda at Maryland, where she can take advantage of UMIACS’ vast computational resources.

A major line of Molloy’s research focuses on the development of phylogeny estimation methods that can effectively utilize distributed-memory systems.

In this context, she says, the genomic data set is distributed across multiple processors, and the algorithm may require these processors to communicate with each other.

“In the worst case, the processors must synchronize with each other at specific points in the algorithm, Molloy explains. “All of this dramatically slows down the computation. My goal is to design methods that reduce communication bottlenecks, while achieving the same accuracy and statistical guarantees of existing methods.”

Molloy says she looks forward to working with graduate students and faculty at Maryland, particularly within the Center for Bioinformatics and Computational Biology.

She has previously collaborated with Mihai Pop, the director of UMIACS, on a project that utilizes estimated phylogenies to perform taxon identification and abundance profiling from metagenomics data sets.

“I hope to continue working with Mihai on problems in metagenomics, where modeling evolutionary processes could prove advantageous,” Molloy says.

Other UMIACS faculty she expects to collaborate with include Brantley Hall, who works on identifying the functions of the genes in the microbiome, and Michael Cummings, who also approaches phylogeny estimation from a high-performance computing lens.

Although Cummings and Molloy utilize different methodologies for phylogeny estimation, Molloy says working together could lead to new approaches.

“There is a lot of potential for creating a hybrid method that combines aspects of our different approaches,” she says. “I look forward to discussing these ideas with my new colleagues at the University of Maryland.”

—Story by Melissa Brachfeld